FPhU5ddlZddlmZddlmZddlmZeejdZd dZ dZ Gdd ejZ dd Z dd Zd Zy)N) check_version)bbox_iouz1.10.0c:|jd}|j\}}}|jdddjdd\}}tj|d|z ||dz fdj|||d} | j dj |S) z Select the positive anchor center in gt. Args: xy_centers (Tensor): shape(h*w, 2) gt_bboxes (Tensor): shape(b, n_boxes, 4) Returns: (Tensor): shape(b, n_boxes, h*w) rrN)dim)shapeviewchunktorchcatamingt_) xy_centers gt_bboxeseps n_anchorsbsn_boxes_ltrb bbox_deltass `C:\Users\daisl\Desktop\realtime-object-detection\venv\Lib\site-packages\ultralytics/utils/tal.pyselect_candidates_in_gtsr s  #I__NB ^^B1 % + +Aq 1FB))Z-2BD9I4IJPQRWWXZ\cenprsK   A  " "3 ''c|jd}|jdkDr|jddkDjd|d}|j d}t j |j|j|j}|jd|jddt j|||j}|jd}|j d}|||fS)aw If an anchor box is assigned to multiple gts, the one with the highest IoI will be selected. Args: mask_pos (Tensor): shape(b, n_max_boxes, h*w) overlaps (Tensor): shape(b, n_max_boxes, h*w) Returns: target_gt_idx (Tensor): shape(b, h*w) fg_mask (Tensor): shape(b, h*w) mask_pos (Tensor): shape(b, n_max_boxes, h*w) rrdtypedevice) summax unsqueezeexpandargmaxrzerosr r#r$scatter_wherefloat)mask_posoverlaps n_max_boxesfg_maskmask_multi_gtsmax_overlaps_idxis_max_overlaps target_gt_idxs rselect_highest_overlapsr6sll2G{{}q!++A.2::2{BO#??1-++hnnHNNS[SbSbc  $4$>$>q$A1E;;~IOOQ,,r"OOB'M '8 ++rcleZdZdZdfd Zej dZdZdZ d dZ dZ xZ S) TaskAlignedAssignerao A task-aligned assigner for object detection. This class assigns ground-truth (gt) objects to anchors based on the task-aligned metric, which combines both classification and localization information. Attributes: topk (int): The number of top candidates to consider. num_classes (int): The number of object classes. alpha (float): The alpha parameter for the classification component of the task-aligned metric. beta (float): The beta parameter for the localization component of the task-aligned metric. eps (float): A small value to prevent division by zero. cvt|||_||_||_||_||_||_y)zJInitialize a TaskAlignedAssigner object with customizable hyperparameters.N)super__init__topk num_classesbg_idxalphabetar)selfr<r=r?r@r __class__s rr;zTaskAlignedAssigner.__init__Ks9  &!   rc|jd|_|jd|_|jdk(r|j}t j |d|j j|t j|j|t j|j|t j|dj|t j|dj|fS|j||||||\}} } t|| |j\} } }|j||| | \} }}| |z} | jdd}| |zjdd}| |z||jzz jdjd}||z}| ||| j| fS)a Compute the task-aligned assignment. Reference https://github.com/Nioolek/PPYOLOE_pytorch/blob/master/ppyoloe/assigner/tal_assigner.py Args: pd_scores (Tensor): shape(bs, num_total_anchors, num_classes) pd_bboxes (Tensor): shape(bs, num_total_anchors, 4) anc_points (Tensor): shape(num_total_anchors, 2) gt_labels (Tensor): shape(bs, n_max_boxes, 1) gt_bboxes (Tensor): shape(bs, n_max_boxes, 4) mask_gt (Tensor): shape(bs, n_max_boxes, 1) Returns: target_labels (Tensor): shape(bs, num_total_anchors) target_bboxes (Tensor): shape(bs, num_total_anchors, 4) target_scores (Tensor): shape(bs, num_total_anchors, num_classes) fg_mask (Tensor): shape(bs, num_total_anchors) target_gt_idx (Tensor): shape(bs, num_total_anchors) rr).rrT)r keepdimr!)sizerr0r$r full_liker>to zeros_like get_pos_maskr6 get_targetsamaxrr'bool)rA pd_scores pd_bboxes anc_points gt_labelsrmask_gtr$r. align_metricr/r5r1 target_labels target_bboxes target_scorespos_align_metrics pos_overlapsnorm_align_metrics rforwardzTaskAlignedAssigner.forwardUs*..#$>>!,   q %%FOOIf$5t{{CFFvNPUP`P`ajPkPnPnouPv$$Y/226:Eioo^g^n^nokk477D,<,ITTU^_I z1-{{7== 9K[K[\ yArr2%**YM]M]^tyy!A  % %b)Aq!AE'M*BD I" !!,"2A6w}}--rctj|jtj|jd}|||j zz}|j j|}|jdd|}|jdtj|jd|jd|jftj|j}|jd|jdd|d d d d d fjdd|j} tj | dkD|d}|||fS) a Compute target labels, target bounding boxes, and target scores for the positive anchor points. Args: gt_labels (Tensor): Ground truth labels of shape (b, max_num_obj, 1), where b is the batch size and max_num_obj is the maximum number of objects. gt_bboxes (Tensor): Ground truth bounding boxes of shape (b, max_num_obj, 4). target_gt_idx (Tensor): Indices of the assigned ground truth objects for positive anchor points, with shape (b, h*w), where h*w is the total number of anchor points. fg_mask (Tensor): A boolean tensor of shape (b, h*w) indicating the positive (foreground) anchor points. Returns: (Tuple[Tensor, Tensor, Tensor]): A tuple containing the following tensors: - target_labels (Tensor): Shape (b, h*w), containing the target labels for positive anchor points. - target_bboxes (Tensor): Shape (b, h*w, 4), containing the target bounding boxes for positive anchor points. - target_scores (Tensor): Shape (b, h*w, num_classes), containing the target scores for positive anchor points, where num_classes is the number of object classes. )rar#r$).Nrrrrr"r N)rrerint64r$r0rdflattenr rgr*r r=r+r'repeatr,) rArPrr5r1 batch_indrSrTrUfg_scores_masks rrJzTaskAlignedAssigner.get_targetss04LLTWWEKK HXHXYZcd % D4D4D(DD !(002=A "r1-m<  Q ]%8%8%;]=P=PQR=SUYUeUe$f*/+++8+?+?A  q-"9"9""=qA At,33Aq$:J:JK NQ$6 qI m]::r) Pg?g@& .>)TN) __name__ __module__ __qualname____doc__r;rno_gradrYrIr\r]rJ __classcell__)rBs@rr8r8<sA U]]_,Z,Z\ 0&*".H-;rr8c gg}}|J|dj|dj}}t|D]\}}||j\} } } } t j | |||z} t j | |||z} t rt j| | dnt j| | \} } |jt j| | fdjdd|jt j| | zdf|||t j|t j|fS) zGenerate anchors from features.r)rar$r#ij)indexingrr rr") r#r$ enumerater rre TORCH_1_10meshgridappendstackr fullr)featsstridesgrid_cell_offset anchor_points stride_tensorr#r$istriderhwsxsys r make_anchorsrs#%r=M   !HNNE!HOO6Ew' 61X^^ 1a \\ae rsR ! 5,,h 7 (&,:z;"))z;z > (ar